This is container_cpu_cfs_throttled_seconds_total
metric. But I'm not sure how to read this graph ?
I can see 27 seconds increment within 08.09.45 to 08.09.52. How this is possible ?
You should wrap this metric into rate() function:
rate(container_cpu_cfs_throttled_seconds_total[5m])
It will show spikes at time ranges when there wasn't enough CPU resources for the container due to CPU limits. The returned non-zero value shows how many additional CPU cores were needed for the container to run without CPU throttling.